CN110599491B - Priori information-based eye image segmentation method, apparatus, device and medium - Google Patents
Priori information-based eye image segmentation method, apparatus, device and medium Download PDFInfo
- Publication number
- CN110599491B CN110599491B CN201910833947.3A CN201910833947A CN110599491B CN 110599491 B CN110599491 B CN 110599491B CN 201910833947 A CN201910833947 A CN 201910833947A CN 110599491 B CN110599491 B CN 110599491B
- Authority
- CN
- China
- Prior art keywords
- image
- segmentation
- loss
- learning model
- loss component
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000003709 image segmentation Methods 0.000 title claims abstract description 86
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000010801 machine learning Methods 0.000 claims abstract description 55
- 238000012549 training Methods 0.000 claims abstract description 51
- 230000011218 segmentation Effects 0.000 claims abstract description 41
- 238000013145 classification model Methods 0.000 claims abstract description 25
- 230000006870 function Effects 0.000 claims description 50
- 238000012545 processing Methods 0.000 claims description 18
- 230000005284 excitation Effects 0.000 claims description 7
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 210000003733 optic disk Anatomy 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 15
- 238000013527 convolutional neural network Methods 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 239000013598 vector Substances 0.000 description 5
- 230000005540 biological transmission Effects 0.000 description 4
- 230000000306 recurrent effect Effects 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 238000013136 deep learning model Methods 0.000 description 3
- 210000003127 knee Anatomy 0.000 description 3
- 238000011176 pooling Methods 0.000 description 3
- 230000004256 retinal image Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000011478 gradient descent method Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000002207 retinal effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30041—Eye; Retina; Ophthalmic
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Eye Examination Apparatus (AREA)
Abstract
The invention discloses an eye image segmentation method, device, equipment and medium based on prior information, wherein the method comprises the steps of acquiring a fundus image dataset and calculating prior information according to the fundus image dataset; constructing a machine learning model, and training the machine learning model based on the prior information to obtain an image segmentation model; acquiring a target image to be segmented, wherein the target image comprises eyes; and inputting the target image into the image segmentation model to obtain a target image segmentation result output by the image segmentation model. According to the method, the prior information is introduced to train the image segmentation model, so that the segmentation result of the target image to be segmented has strong interpretation and accuracy, the classification model can be trained based on the prior information, and the classification model is used for classifying the segmentation result of the target image to be segmented, so that the classification result with strong interpretation and high accuracy is obtained, and the scheme of the method can have wider application prospect.
Description
Technical Field
The present invention relates to the field of image processing, and in particular, to a method, apparatus, device, and medium for segmenting an eye image based on prior information.
Background
In the prior art, there are two main schemes for classifying eye images based on machine learning, namely, the first scheme is to classify eye images based on a deep learning model MRNet, and the second scheme is to classify eye images based on a cyclic attention convolutional neural network (Recurrent Attention Convolutional Neural Network).
As shown in fig. 1 (a), a schematic structural diagram of a deep learning model MRNet is shown, the MRNet is based on a convolutional neural network alexnealexet structure, a corresponding two-dimensional image of a three-dimensional knee image is taken as an input, and the probability that the three-dimensional knee image points to a certain target object can be finally given by the output of different convolutional layers through a maximum pooling operation. However, the training process of MRNet does not introduce a priori knowledge, so that it cannot explain the probability that the three-dimensional knee image points to a certain target object.
As shown in fig. 1 (b), which shows a schematic structural diagram of Recurrent Attention Convolutional Neural Network, recurrent Attention Convolutional Neural Network is a natural image based classification technique. By introducing an attention branch, attention weight is introduced when the characteristics are extracted, but training of the attention branch is not limited by priori knowledge, and the risk of errors of the attention area exists, so that the accuracy of the attention branch is still to be improved.
In summary, the prior knowledge is difficult to use in both eye image classification methods, so that the problem of poor interpretability and poor directivity of the image classification result exists, and the application prospect of the fundus image segmentation result is limited.
Disclosure of Invention
In order to solve the technical problem that prior knowledge is not used for eye image classification in the prior art, so that image segmentation and classification results have poor interpretation, the embodiment of the invention provides an eye image segmentation method, device, equipment and medium based on prior information.
In one aspect, the present invention provides a method for segmenting an eye image based on prior information, the method comprising:
acquiring a fundus image dataset, and calculating prior information according to the fundus image dataset;
constructing a machine learning model, and training the machine learning model based on the prior information to obtain an image segmentation model;
acquiring a target image to be segmented, wherein the target image comprises eyes;
and inputting the target image into the image segmentation model to obtain a target image segmentation result output by the image segmentation model.
In another aspect, the present invention provides an eye image segmentation apparatus based on prior information, the apparatus comprising:
the fundus image data set acquisition module is used for acquiring a fundus image data set and calculating prior information according to the fundus image data set;
the machine learning model training module is used for constructing a machine learning model, and training the machine learning model based on the prior information to obtain an image segmentation model;
the device comprises a target image acquisition module, a target image segmentation module and a target image segmentation module, wherein the target image acquisition module is used for acquiring a target image to be segmented, and the target image comprises eyes;
and the segmentation module is used for inputting the target image into the image segmentation model to obtain a target image segmentation result output by the image segmentation model.
In another aspect, the present invention provides an eye image segmentation apparatus based on prior information, which is characterized in that the apparatus includes a processor and a memory, where at least one instruction, at least one program, a code set, or an instruction set is stored in the memory, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement an eye image segmentation method based on prior information.
In another aspect, the present invention provides a computer storage medium, wherein at least one instruction, at least one program, a code set, or an instruction set is stored in the storage medium, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded by a processor and executes an eye image segmentation method based on prior information.
The invention provides an eye image segmentation method, device, equipment and medium based on priori information. According to the invention, the image segmentation model is trained by introducing priori information, so that the segmentation result of the target image to be segmented has stronger interpretation and accuracy. Correspondingly, the embodiment of the invention can further train a classification model based on prior information, and classify the segmentation result of the target image to be segmented by using the classification model so as to obtain a classification result with strong interpretability and high accuracy. Different from the prior art, the embodiment of the invention introduces priori knowledge to train the image segmentation model and the classification model, so that the image segmentation result and the classification result obtained by the embodiment of the invention have interpretability, and the scheme of the embodiment of the invention can have wider application prospect.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
Fig. 1 (a) is a schematic structural diagram of a deep learning model MRNet provided by the invention;
fig. 1 (b) is a schematic structural diagram of Recurrent Attention Convolutional Neural Network provided by the present invention;
fig. 2 is a schematic view of an implementation environment of an eye image segmentation method based on prior information provided by the invention;
FIG. 3 is a flow chart of an eye image segmentation method based on prior information provided by the invention;
FIG. 4 is a flow chart of a prior information calculation from the fundus image dataset provided by the present invention;
FIG. 5 is a schematic diagram of a machine learning model provided by the present invention;
FIG. 6 is a block diagram of a self-cleaving network architecture provided by the present invention;
FIG. 7 is a schematic diagram of data processing logic of the self-segmentation network provided by the present invention;
FIG. 8 is a flow chart of a method for constructing a loss function according to the present invention;
FIG. 9 is a flow chart of the build center loss component provided by the present invention;
FIG. 10 is a schematic diagram of a joint learning model provided by the invention, which is obtained by combining a machine learning model and a classification learning model;
FIG. 11 is a block diagram of an eye image segmentation device based on prior information provided by the invention;
fig. 12 is a schematic hardware structure of an apparatus for implementing the method provided by the embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or server that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the detailed description and specific examples, while indicating the embodiment of the invention, are intended for purposes of illustration only and are not intended to limit the scope of the invention.
The terms "first" and "second" are used below for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. In the description of the present embodiment, unless otherwise specified, the meaning of "plurality" is two or more.
The embodiment of the invention provides an eye image segmentation method based on prior information. Firstly, the embodiment of the invention discloses an implementation environment of the eye image segmentation method based on prior information in a feasible embodiment.
Referring to fig. 2, the implementation environment includes: client 01, server 03.
The client 01 may include: the smart phone, tablet computer, notebook computer, digital assistant, intelligent wearable device, vehicle terminal and other type of physical devices can also include software running in the physical devices, such as application programs with eye image segmentation function. The Client 01 may be communicatively connected to the Server 03 based on Browser/Server (B/S) or Client/Server (C/S) mode.
The client 01 may send a target image to be segmented to the server 03, where the target image includes an eye image, the server 03 may train an image segmentation model based on preset prior information, output a segmentation result of the target image according to the image segmentation model, and transmit the segmentation result to the client 01. In a preferred embodiment, the server 03 may further classify the target image according to the segmentation result of the target image, specifically, may train an image classification model, obtain a classification result by inputting the segmentation result into the target classification model, and transmit the classification result to the client 01. For example, the classification result may be expressed as a probability that the target image belongs to a target class.
The server 03 may comprise a server running independently, or a distributed server, or a server cluster consisting of a plurality of servers.
Referring to fig. 3, a flowchart of an eye image segmentation method based on prior information is shown, where the method may be implemented by using a server in the implementation environment shown in fig. 2 as an execution subject, and the method may include:
s101, acquiring a fundus image dataset, and calculating prior information according to the fundus image dataset.
In one possible embodiment, the fundus image dataset may use existing retinal image data (FIRE: fundus Image Registration Dataset). Further, the retinal image data may also serve as a data source for a training set for subsequent training of the machine learning model.
FIRE is a retinal fundus image dataset containing 129 fundus retinal images, combined 134 by different feature combinations, and including labels for classification of the images with their corresponding target objects. These image combinations are classified into 3 classes according to the characteristics. Fundus images were acquired by a Nidek AFC-210 fundus camera with a resolution of 2912x2912 and a visual elevation angle of 40 degrees. Images were co-constructed by Papageorgiou Hospital hospital and Aristotle University of Thessaloniki university, collected from 39 users by the university of Thessaloniki.
Specifically, in the embodiment of the present invention, the prior information is calculated according to the fundus image dataset, as shown in fig. 4, including:
s1011, obtaining a first segmentation image set, a second segmentation image set and a fundus pit central position coordinate point set according to the fundus image data set, wherein the first segmentation image set comprises a plurality of fundus optic disc segmentation images, and the second segmentation image set comprises a plurality of fundus optic cup segmentation images.
S1013, calculating a space probability distribution result of the fundus optic disk position according to the first segmentation image set, calculating a space probability distribution result of the fundus optic cup position according to the second segmentation image set, and calculating a space probability distribution result of the fundus fovea center position according to the fundus fovea center position coordinate point set.
S103, constructing a machine learning model, and training the machine learning model based on the prior information to obtain an image segmentation model.
Specifically, as shown in fig. 5, the machine learning model includes a feature extractor, a self-separating network, and a semantic component basis, wherein the feature extractor may use a pre-trained neural network, and the self-separating network and the semantic component basis are both training objects of the machine learning model. The images in the training set are transmitted to the feature extractor and the self-separating network, respectively, to facilitate training the self-separating network and the semantic component basis. Specifically, the feature extractor performs feature extraction on the images in the training set to obtain a first feature atlas, the self-segmentation network segments the images in the training set to obtain a mask set of segmented images, the mask set of segmented images and the semantic component basis are processed through a preset excitation function to obtain a second feature atlas, and the difference value obtained by the first feature atlas and the second feature atlas points to a first loss function of the machine learning model. And training the machine learning model based on the first loss function, and taking a self-separation network in the trained machine learning model as an image segmentation model. And the mask set of the segmented image output by the self-segmentation network can obtain the segmented image corresponding to the original input image through the dot multiplication operation of the mask set and the original input image.
In one possible embodiment, the excitation function is represented as a linear rectification function (Rectified Linear Unit, reLU), which is an activation function (activation function) commonly used in the machine learning art, generally referring to a nonlinear function represented by a ramp function and its variants. In general terms, a linear rectification function refers to a ramp function in mathematics, i.e., f (x) =max (0, x), whereas in the machine learning field, linear rectification is used as an activation function of a machine learning model, which defines a nonlinear output result of the machine learning model after linear transformation.
In particular, the feature extractor may use a pre-trained VGG network. VGG networks, collectively Visual Geometry Group, belong to the scientific engineering line of the university of oxford, and are deep convolutional neural networks developed by researchers of the university of oxford, visual geometry group (Visual Geometry Group) and Google deep Mind. The convolutional network model comprises a plurality of columns of convolutional network models beginning with VGG, and can be applied to face recognition, image classification and the like from VGG16 to VGG19 respectively. VGG networks employ a small convolution kernel of 3x3 at all layers while deepening the number of network layers in order to avoid excessive parameters, the convolution layer step size being set to 1. The input to the VGG is set to an RGB image of 224x244 size, the RGB average is calculated over all images on the training set image, then the images are passed as input into the VGG network, the convolution step size is fixed by 1 using a convolution kernel of 3x3 or 1x 1. The VGG network full-connection layer has 3 layers, VGG11 can be obtained according to the difference of total number of convolution layers and full-connection layers until VGG19, the minimum VGG11 has 8 convolution layers and 3 full-connection layers, and the maximum VGG19 has 16 convolution layers and +3 full-connection layers. The embodiments of the present invention are not limited to specific branches of the VGG network, for example, any one of VGGs 16 to 19 may be used. When the VGG network is initialized, the weight of the VGG network can be pre-trained by adopting the natural image.
In order to obtain a better image segmentation effect, the embodiment of the present invention should be a feasible self-segmentation network structure, as shown in fig. 6, where the self-segmentation network sequentially includes a first convolution layer, a first residual layer, a first maximum pooling layer, a second residual layer, a second maximum pooling layer, a residual combination layer, a first deconvolution layer, a second convolution layer, a second deconvolution layer, a convolution combination layer and a linear interpolation layer along a data processing sequence, where the residual combination layer includes three continuous residual layers, and the convolution combination layer includes two continuous convolution layers.
Referring to fig. 7, a schematic diagram of the data processing logic of the self-segmentation network is shown, where each rectangle in the schematic diagram represents a feature map after corresponding data processing.
As shown in fig. 5, the machine learning model includes three branches, a first branch is used for acquiring images in a training set, the images are input into a feature extractor, and a first feature atlas is obtained according to the output of the feature extractor under the condition of a preset significance constraint (saliency constraint). Specifically, in the embodiment of the present invention, the first feature atlas may use C x H x W x And (C) represents the number of channels of the feature extractor, the length of the two-dimensional feature map, the width of the two-dimensional feature map and the channel index, respectively.
The second branch is used for acquiring images in the training set, inputting the images into a self-segmentation network to obtain a mask set of segmented images, and correspondingly, the mask set of the segmented images generated by the second branch can use K y H y W y The representation, K, H, W, y, represents the number of channels from the self-separating network, the length, width of the mask, and the channel subscripts, respectively.
The third branch is used for outputting semantic component bases matched with the segmented image set, the semantic component bases generate K vectors, and the length of each vector is C.
In order to embody the influence of priori knowledge on the training of the machine learning model in the embodiment of the invention, the trained image segmentation model can segment the target image based on the guidance of the priori knowledge, so that a segmentation result with strong interpretability and high precision is obtained.
In a possible embodiment, the embodiment of the present invention provides a first loss function construction method, as shown in fig. 8, where the method includes:
s1, constructing a central loss component, wherein the central loss component points to the central loss generated in the second characteristic diagram set.
Specifically, the construction center loss component, as shown in fig. 9, includes:
s11, acquiring characteristic images of the corresponding channels in the second characteristic image set.
S13, calculating the mass center position of each characteristic image.
Specifically, the centroid position calculation formula isWhere y is the channel label and R (y, u, v) represents the value of the feature space of the feature image at (u, v) in the y-th channel.
S15, calculating the center loss component according to the centroid positions of the characteristic images.
Specifically, the center loss component calculation formula is
S3, constructing a semantic loss component, wherein the semantic loss component points to loss generated by difference between the characteristic images in the first characteristic image set and the characteristic images in the second characteristic image set corresponding to the semantic loss component.
Specifically, the semantic loss component calculation formula is thatWherein v (u, v) points to all feature images in said first feature map set,/v>Pointing to all feature images, w, in the second feature map set y Vector generated pointing to semantic component basis.
S5, constructing an orthogonal loss component, wherein the orthogonal loss component points to the degree of orthogonality of vectors generated by the semantic component basis.
Specifically, the orthogonal loss component calculation formula is L on ＝||WW T -I k || F 2 Wherein W is a matrix formed by vectors with K length C generated by the semantic component basis, I k Is WW (world wide web) T And (5) an adaptive identity matrix.
S7, constructing a priori loss component, wherein the priori loss component points to the degree of deviation between the second feature atlas and the priori information.
In a possible embodiment, if the prior information includes a spatial probability distribution result of the fundus optic disc position, a spatial probability distribution result of the fundus cup position, and a spatial probability distribution result of the fundus fovea center position, the prior loss component may be represented by using a mean square value of distances from the fundus optic disc position, the fundus cup position, and the fundus fovea center position of the prior information, respectively, in the second feature map set.
In another possible embodiment, the formula is based onCalculating a priori loss component, where R k ,P k And respectively representing the positions of the position points in the second characteristic diagram set and the positions of the maximum probability points corresponding to the position points determined according to the prior information.
S9, constructing a first loss function according to the central loss component, the semantic loss component, the orthogonal loss component and the priori loss component.
In particular, the first loss function, L, can be constructed based on a weighting method 1 ＝λ con L con +λ SC L SC +λ on L on +λ mse L mse Wherein lambda is con ,λ SC ,λ on ,λ mse Is the weight.
After determining the structure of the machine learning model and the first loss function, the machine learning model may be trained according to a gradient descent method. And (3) accurately segmenting the target image according to an image segmentation model obtained by the machine learning model obtained after training.
In a preferred embodiment, the dataset used to train the machine learning model may also include labels, such as FIRE, for the classification of the image and its corresponding target object. Accordingly, a classification learning model can be designed, and the machine learning model and the classification learning model are trained in a combined manner, so that an image segmentation model and a classification model can be obtained together.
In particular, the classification learning model may use a Resnet10 model, resnet10 being a deep residual network. The depth residual network introduces a residual network structure, and by the residual network structure, the network depth can be greatly enhanced, and a better classification effect can be obtained. The residual network structure has a hopping structure. The residual network structure uses the thought of cross-layer linking of the high-speed network, but the residual item adopts identity mapping.
As shown in fig. 10, a schematic diagram of a joint learning model obtained by combining a machine learning model with a classification learning model is shown, the joint learning model takes an image in a training set as an input, a mask set of a segmented image is output from a segmentation network, the mask set of the segmented image and image points in the training set are multiplied to obtain a segmented image set, and labels of the segmented image set and images in the corresponding training set are taken as the input of the classification learning model.
The loss component generated by the class learning model may be characterized by a cross entropy loss function, which may be expressed asWherein y, & gt>Respectively pointing to target objects output by the training set labels and the classification learning model.
Correspondingly, the second loss function of the joint learning model can be constructed based on the first loss function, and only the loss component generated by the classification learning model and the corresponding weight are added on the basis of the first loss function, namely the second loss function can be expressed as L 2 ＝λ con L con +λ SC L SC +λ on L on +λ mse L mse +λ c L c Wherein lambda is c The weight of the penalty generated for classifying the learning model. Of course, the joint learning model is also trained using a gradient descent method. And taking the self-segmentation network and the classification learning model in the trained joint learning model as an image segmentation model and a classification model respectively.
S105, acquiring a target image to be segmented, wherein the target image comprises eyes.
S107, inputting the target image into the image segmentation model to obtain a target image segmentation result output by the image segmentation model.
Specifically, the image segmentation model may obtain a mask set of a segmented image, and the target segmentation result may be obtained by dot multiplying the mask set of the segmented image with the target image.
In a preferred embodiment, the image segmentation model and the classification model can be obtained by constructing a joint learning model and training the joint learning model, so after obtaining the target segmentation result, the method further comprises the following steps:
s109, inputting the target image segmentation result into a classification model to obtain a classification result output by the classification model, wherein the classification model is obtained by training a joint learning model based on the prior information, and the joint learning model comprises the machine learning model and the classification learning model.
The embodiment of the invention discloses an eye image segmentation method based on priori information, which is characterized in that an image segmentation model is trained by introducing the priori information, so that the segmentation result of a target image to be segmented has stronger interpretation and accuracy. Correspondingly, the embodiment of the invention can further train a classification model based on prior information, and classify the segmentation result of the target image to be segmented by using the classification model so as to obtain a classification result with strong interpretability and high accuracy. Different from the prior art, the embodiment of the invention introduces priori knowledge to train the image segmentation model and the classification model, so that the image segmentation result and the classification result obtained by the embodiment of the invention have interpretability, and the scheme of the embodiment of the invention can have wider application prospect.
The embodiment of the invention also discloses an eye image segmentation device based on prior information, as shown in fig. 11, the device comprises:
a fundus image dataset acquisition module 201 for acquiring a fundus image dataset from which a priori information is calculated.
A machine learning model training module 203, configured to construct a machine learning model, and train the machine learning model based on the prior information to obtain an image segmentation model.
The machine learning model comprises a feature extractor, a self-segmentation network and a semantic component base, wherein the feature extractor can use a pre-trained neural network, and the self-segmentation network and the semantic component base are training objects of the machine learning model. The images in the training set are transmitted to the feature extractor and the self-separating network, respectively, to facilitate training the self-separating network and the semantic component basis. Specifically, the feature extractor performs feature extraction on the images in the training set to obtain a first feature atlas, the self-segmentation network segments the images in the training set to obtain a mask set of segmented images, the mask set of segmented images and the semantic component basis are processed through a preset excitation function to obtain a second feature atlas, and the difference value obtained by the first feature atlas and the second feature atlas points to a first loss function of the machine learning model. And training the machine learning model based on the first loss function, and taking a self-separation network in the trained machine learning model as an image segmentation model. And the mask set of the segmented image output by the self-segmentation network can obtain the segmented image corresponding to the original input image through the dot multiplication operation of the mask set and the original input image.
A target image acquisition module 205, configured to acquire a target image to be segmented, where the target image includes an eye.
The segmentation module 207 is configured to input the target image to the image segmentation model to obtain a target image segmentation result output by the image segmentation model.
Further, the method may further include:
a classification module 209 inputs the target image segmentation result into a classification model to obtain a classification result output by the classification model, the classification model being obtained by training a joint learning model based on the prior information, the joint learning model including the machine learning model and a classification learning model.
The joint learning model takes images in a training set as input, a mask set of a segmented image is output by a self-segmentation network, the mask set of the segmented image and image points in the training set are multiplied to obtain a segmented image set, and labels of the images in the segmented image set and the corresponding training set are taken as input of the classification learning model.
Specifically, the embodiments of the eye image segmentation device and method based on prior information according to the embodiments of the present invention are all based on the same inventive concept. Please refer to the method embodiment for details, which will not be described herein.
The embodiment of the invention also provides a computer storage medium which can store a plurality of instructions. The instructions may be adapted to be loaded and executed by a processor to perform a method for segmenting an eye image based on a priori information according to an embodiment of the invention, the method at least comprising the steps of:
an eye image segmentation method based on prior information, the method comprising:
acquiring a fundus image dataset, and calculating prior information according to the fundus image dataset;
constructing a machine learning model, and training the machine learning model based on the prior information to obtain an image segmentation model;
acquiring a target image to be segmented, wherein the target image comprises eyes;
and inputting the target image into the image segmentation model to obtain a target image segmentation result output by the image segmentation model.
In a preferred embodiment, further comprising:
inputting the target image segmentation result into a classification model to obtain a classification result output by the classification model, wherein the classification model is obtained by training a joint learning model based on the prior information, and the joint learning model comprises the machine learning model and the classification learning model.
In a preferred embodiment, the machine learning model includes three branches, a first branch is used for acquiring images in a training set, the images are input into a feature extractor, and a first feature atlas is obtained according to the output of the feature extractor under a preset significance constraint condition;
the second branch is used for acquiring images in a training set, and inputting the images into a self-segmentation network to obtain a mask set for segmenting the images;
the third branch is used for outputting semantic component bases matched with the segmented image set;
and processing the mask set of the segmented image and the semantic component base through a preset excitation function to obtain a second feature atlas, wherein the difference value obtained by the first feature atlas and the second feature atlas points to a first loss function of the machine learning model.
In a preferred embodiment, the method further comprises the step of constructing a first loss function, said constructing a first loss function comprising:
constructing a center loss component, the center loss component pointing to a center loss generated in the second feature atlas;
constructing a semantic loss component, wherein the semantic loss component points to loss generated by difference between a characteristic image in a first characteristic image set and a characteristic image in a second characteristic image set corresponding to the semantic loss component;
constructing a priori loss component, wherein the priori loss component points to the degree of deviation between the second feature atlas and the priori information;
a first loss function is constructed from the center loss component, the semantic loss component, the quadrature loss component, and the prior loss component.
In a preferred embodiment, the method further comprises the step of constructing a second loss function, training a joint learning model based on the second loss function, the constructing the second loss function comprising:
constructing a loss component generated by a classification learning model;
and constructing a second loss function number according to the central loss component, the semantic loss component, the orthogonal loss component, the prior loss component and the loss component generated by the classification learning model.
In a preferred embodiment, said constructing a center loss component comprises:
acquiring characteristic images of the corresponding channels in the second characteristic image set;
calculating the centroid position of each characteristic image;
the center loss component is calculated from the centroid positions of the respective feature images.
In a preferred embodiment, said calculating a priori information from said fundus image dataset comprises:
obtaining a first segmentation image set, a second segmentation image set and a fundus pit central position coordinate point set according to the fundus image data set, wherein the first segmentation image set comprises a plurality of fundus optic disc segmentation images, and the second segmentation image set comprises a plurality of fundus optic cup segmentation images;
and calculating a space probability distribution result of the fundus optic disk position according to the first segmentation image set, calculating a space probability distribution result of the fundus optic cup position according to the second segmentation image set, and calculating a space probability distribution result of the fundus fovea center position according to the fundus fovea center position coordinate point set.
Further, fig. 12 shows a schematic diagram of a hardware structure of an apparatus for implementing the method provided by the embodiment of the present invention, where the apparatus may participate in forming or including the device or the system provided by the embodiment of the present invention. As shown in fig. 12, the apparatus 10 may include one or more processors 102 (shown as 102a, 102b, … …,102 n) which may include, but are not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA, a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, the method may further include: a display, an input/output interface (I/O interface), a Universal Serial Bus (USB) port (which may be included as one of the ports of the I/O interface), a network interface, a power supply, and/or a camera. It will be appreciated by those of ordinary skill in the art that the configuration shown in fig. 12 is merely illustrative and is not intended to limit the configuration of the electronic device described above. For example, the device 10 may also include more or fewer components than shown in fig. 12, or have a different configuration than shown in fig. 12.
It should be noted that the one or more processors 102 and/or other data processing circuits described above may be referred to generally herein as "data processing circuits. The data processing circuit may be embodied in whole or in part in software, hardware, firmware, or any other combination. Further, the data processing circuitry may be a single stand-alone processing module, or incorporated in whole or in part into any of the other elements in the device 10 (or mobile device). As referred to in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of the path of the variable resistor termination to interface).
The memory 104 may be used to store software programs and modules of application software, and the processor 102 executes the software programs and modules stored in the memory 104 to perform various functional applications and data processing, i.e. implement an eye image segmentation method based on a priori information as described above. Memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, memory 104 may further include memory located remotely from processor 102, which may be connected to device 10 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The transmission means 106 is arranged to receive or transmit data via a network. Specific examples of the network described above may include a wireless network provided by a communications provider of device 10. In one example, the transmission device 106 includes a network adapter (NetworkInterfaceController, NIC) that can connect to other network devices through a base station to communicate with the internet. In one example, the transmission device 106 may be a radio frequency (RadioFrequency, RF) module for communicating wirelessly with the internet.
The display may be, for example, a touch screen type Liquid Crystal Display (LCD) that may enable a user to interact with a user interface of the device 10 (or mobile device).
It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. And the foregoing description has been directed to specific embodiments of this specification. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for the device and server embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, and references to the parts of the description of the method embodiments are only required.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program for instructing relevant hardware, where the program may be stored in a computer readable storage medium, and the storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
The foregoing description of the preferred embodiments of the invention is not intended to limit the invention to the precise form disclosed, and any such modifications, equivalents, and alternatives falling within the spirit and scope of the invention are intended to be included within the scope of the invention.
Claims (9)
1. An eye image segmentation method based on prior information, which is characterized by comprising the following steps:
acquiring a fundus image dataset, and calculating prior information according to the fundus image dataset;
constructing a machine learning model, and training the machine learning model based on the prior information to obtain an image segmentation model; the first loss function of the machine learning model is obtained based on a first feature atlas and a second feature atlas, and the first feature atlas is obtained by extracting features from images in a training set under a preset significance constraint condition; the second feature atlas is obtained by processing a mask set of the segmented image corresponding to the image and an adaptive semantic component base through an excitation function;
acquiring a target image to be segmented, wherein the target image comprises eyes;
inputting the target image into the image segmentation model to obtain a target image segmentation result output by the image segmentation model;
wherein the first loss function is constructed by:
constructing a center loss component, the center loss component pointing to a center loss generated in the second feature atlas;
constructing a semantic loss component, wherein the semantic loss component points to loss generated by difference between a characteristic image in a first characteristic image set and a characteristic image in a second characteristic image set corresponding to the semantic loss component;
constructing a priori loss component, wherein the priori loss component points to the degree of deviation between the second feature atlas and the priori information;
a first loss function is constructed from the center loss component, the semantic loss component, the quadrature loss component, and the prior loss component.
2. The method as recited in claim 1, further comprising:
inputting the target image segmentation result into a classification model to obtain a classification result output by the classification model, wherein the classification model is obtained by training a joint learning model based on the prior information, and the joint learning model comprises the machine learning model and the classification learning model.
3. The method according to claim 1 or 2, characterized in that:
the machine learning model comprises three branches, wherein the first branch is used for acquiring images in the training set, inputting the images into a feature extractor, and obtaining the first feature atlas according to the output of the feature extractor under the preset significance constraint condition;
the second branch is used for acquiring images in the training set, and inputting the images into a self-dividing network to obtain a mask set of the divided images;
a third branch for outputting the semantic component basis adapted to the segmented image set;
and processing the mask set of the segmented image and the semantic component basis through a preset excitation function to obtain a second feature atlas, wherein the difference value obtained by the first feature atlas and the second feature atlas points to a first loss function of the machine learning model.
4. A method according to claim 3, further comprising the step of constructing a second loss function, training a joint learning model based on the second loss function, the constructing a second loss function comprising:
constructing a loss component generated by a classification learning model;
and constructing a second loss function according to the central loss component, the semantic loss component, the orthogonal loss component, the prior loss component and the loss component generated by the classification learning model.
5. The method of claim 4, wherein the constructing a center loss component comprises:
acquiring characteristic images of the corresponding channels in the second characteristic image set;
calculating the centroid position of each characteristic image;
the center loss component is calculated from the centroid position of each of the feature images.
6. The method according to claim 1 or 2, wherein said calculating a priori information from said fundus image dataset comprises:
obtaining a first segmentation image set, a second segmentation image set and a fundus pit central position coordinate point set according to the fundus image data set, wherein the first segmentation image set comprises a plurality of fundus optic disc segmentation images, and the second segmentation image set comprises a plurality of fundus optic cup segmentation images;
and calculating a space probability distribution result of the fundus optic disk position according to the first segmentation image set, calculating a space probability distribution result of the fundus optic cup position according to the second segmentation image set, and calculating a space probability distribution result of the fundus fovea center position according to the fundus fovea center position coordinate point set.
7. An ocular image segmentation device based on a priori information, the device comprising:
the fundus image data set acquisition module is used for acquiring a fundus image data set and calculating prior information according to the fundus image data set;
the machine learning model training module is used for constructing a machine learning model, and training the machine learning model based on the prior information to obtain an image segmentation model; the first loss function of the machine learning model is obtained based on a first feature atlas and a second feature atlas, and the first feature atlas is obtained by extracting features from images in a training set under a preset significance constraint condition; the second feature atlas is obtained by processing a mask set of the segmented image corresponding to the image and an adaptive semantic component base through an excitation function;
the device comprises a target image acquisition module, a target image segmentation module and a target image segmentation module, wherein the target image acquisition module is used for acquiring a target image to be segmented, and the target image comprises eyes;
the segmentation module is used for inputting the target image into the image segmentation model to obtain a target image segmentation result output by the image segmentation model;
wherein the first loss function is constructed by:
constructing a center loss component, the center loss component pointing to a center loss generated in the second feature atlas;
constructing a semantic loss component, wherein the semantic loss component points to loss generated by difference between a characteristic image in a first characteristic image set and a characteristic image in a second characteristic image set corresponding to the semantic loss component;
constructing a priori loss component, wherein the priori loss component points to the degree of deviation between the second feature atlas and the priori information;
a first loss function is constructed from the center loss component, the semantic loss component, the quadrature loss component, and the prior loss component.
8. A computer storage medium having stored therein at least one instruction, at least one program, code set, or instruction set, loaded and executed by a processor to implement a prior information based ocular image segmentation method according to any of claims 1-6.
9. An ophthalmic image segmentation device based on a priori information, characterized in that it comprises a processor and a memory in which at least one instruction, at least one program, code set or instruction set is stored, said at least one instruction, said at least one program, said code set or instruction set being loaded by said processor and executing an ophthalmic image segmentation method based on a priori information as claimed in any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910833947.3A CN110599491B (en) | 2019-09-04 | 2019-09-04 | Priori information-based eye image segmentation method, apparatus, device and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910833947.3A CN110599491B (en) | 2019-09-04 | 2019-09-04 | Priori information-based eye image segmentation method, apparatus, device and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110599491A CN110599491A (en) | 2019-12-20 |
| CN110599491B true CN110599491B (en) | 2024-04-12 |
Family
ID=68857447
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910833947.3A Active CN110599491B (en) | 2019-09-04 | 2019-09-04 | Priori information-based eye image segmentation method, apparatus, device and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110599491B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111209854A (en) * | 2020-01-06 | 2020-05-29 | 苏州科达科技股份有限公司 | Method and device for recognizing unbelted driver and passenger and storage medium |
| CN111598904B (en) * | 2020-05-21 | 2022-09-16 | 腾讯科技（深圳）有限公司 | Image segmentation method, device, equipment and storage medium |
| CN112052721A (en) * | 2020-07-16 | 2020-12-08 | 北京邮电大学 | Wink oscillogram generation method, device and equipment based on deep learning |
| CN112287938B (en) * | 2020-10-29 | 2022-12-06 | 苏州浪潮智能科技有限公司 | Text segmentation method, system, device and medium |
| CN112434576A (en) * | 2020-11-12 | 2021-03-02 | 合肥的卢深视科技有限公司 | Face recognition method and system based on depth camera |
| CN112598686B (en) * | 2021-03-03 | 2021-06-04 | 腾讯科技（深圳）有限公司 | Image segmentation method and device, computer equipment and storage medium |
| CN113658187B (en) * | 2021-07-26 | 2024-03-29 | 南方科技大学 | Medical image segmentation method, device and storage medium combined with anatomical priori |
| CN114742987B (en) * | 2022-06-08 | 2022-09-27 | 苏州市洛肯电子科技有限公司 | Automatic positioning control method and system for cutting of non-metallic materials |
| CN117132777B (en) * | 2023-10-26 | 2024-03-22 | 腾讯科技（深圳）有限公司 | Image segmentation method, device, electronic equipment and storage medium |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104318565A (en) * | 2014-10-24 | 2015-01-28 | 中南大学 | Interactive method for retinal vessel segmentation based on bidirectional region growing of constant-gradient distance |
| CN108717868A (en) * | 2018-04-26 | 2018-10-30 | 博众精工科技股份有限公司 | Glaucoma eye fundus image screening method based on deep learning and system |
| CN108921227A (en) * | 2018-07-11 | 2018-11-30 | 广东技术师范学院 | A kind of glaucoma medical image classification method based on capsule theory |
| CN109145939A (en) * | 2018-07-02 | 2019-01-04 | 南京师范大学 | A kind of binary channels convolutional neural networks semantic segmentation method of Small object sensitivity |
| CN109598733A (en) * | 2017-12-31 | 2019-04-09 | 南京航空航天大学 | Retinal fundus images dividing method based on the full convolutional neural networks of depth |
| CN109635862A (en) * | 2018-12-05 | 2019-04-16 | 合肥奥比斯科技有限公司 | Retinopathy of prematurity plus lesion classification method |
| CN109658423A (en) * | 2018-12-07 | 2019-04-19 | 中南大学 | A kind of optic disk optic cup automatic division method of colour eyeground figure |
| CN109829877A (en) * | 2018-09-20 | 2019-05-31 | 中南大学 | A kind of retinal fundus images cup disc ratio automatic evaluation method |
| CN109886965A (en) * | 2019-04-09 | 2019-06-14 | 山东师范大学 | The layer of retina dividing method and system that a kind of level set and deep learning combine |
| CN109949302A (en) * | 2019-03-27 | 2019-06-28 | 天津工业大学 | Retinal feature Structural Techniques based on pixel |
| CN109978893A (en) * | 2019-03-26 | 2019-07-05 | 腾讯科技（深圳）有限公司 | Training method, device, equipment and the storage medium of image, semantic segmentation network |
| CN110189327A (en) * | 2019-04-15 | 2019-08-30 | 浙江工业大学 | Eye ground blood vessel segmentation method based on structuring random forest encoder |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3884844A1 (en) * | 2008-07-18 | 2021-09-29 | Doheny Eye Institute | Optical coherence tomography-based ophthalmic testing methods, devices and systems |
| WO2011059409A1 (en) * | 2009-11-16 | 2011-05-19 | Jiang Liu | Obtaining data for automatic glaucoma screening, and screening and diagnostic techniques and systems using the data |
| JP5631032B2 (en) * | 2010-03-19 | 2014-11-26 | キヤノン株式会社 | Image processing apparatus, image processing system, image processing method, and program for causing computer to execute image processing |
| US9779492B1 (en) * | 2016-03-15 | 2017-10-03 | International Business Machines Corporation | Retinal image quality assessment, error identification and automatic quality correction |
-
2019
- 2019-09-04 CN CN201910833947.3A patent/CN110599491B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104318565A (en) * | 2014-10-24 | 2015-01-28 | 中南大学 | Interactive method for retinal vessel segmentation based on bidirectional region growing of constant-gradient distance |
| CN109598733A (en) * | 2017-12-31 | 2019-04-09 | 南京航空航天大学 | Retinal fundus images dividing method based on the full convolutional neural networks of depth |
| CN108717868A (en) * | 2018-04-26 | 2018-10-30 | 博众精工科技股份有限公司 | Glaucoma eye fundus image screening method based on deep learning and system |
| CN109145939A (en) * | 2018-07-02 | 2019-01-04 | 南京师范大学 | A kind of binary channels convolutional neural networks semantic segmentation method of Small object sensitivity |
| CN108921227A (en) * | 2018-07-11 | 2018-11-30 | 广东技术师范学院 | A kind of glaucoma medical image classification method based on capsule theory |
| CN109829877A (en) * | 2018-09-20 | 2019-05-31 | 中南大学 | A kind of retinal fundus images cup disc ratio automatic evaluation method |
| CN109635862A (en) * | 2018-12-05 | 2019-04-16 | 合肥奥比斯科技有限公司 | Retinopathy of prematurity plus lesion classification method |
| CN109658423A (en) * | 2018-12-07 | 2019-04-19 | 中南大学 | A kind of optic disk optic cup automatic division method of colour eyeground figure |
| CN109978893A (en) * | 2019-03-26 | 2019-07-05 | 腾讯科技（深圳）有限公司 | Training method, device, equipment and the storage medium of image, semantic segmentation network |
| CN109949302A (en) * | 2019-03-27 | 2019-06-28 | 天津工业大学 | Retinal feature Structural Techniques based on pixel |
| CN109886965A (en) * | 2019-04-09 | 2019-06-14 | 山东师范大学 | The layer of retina dividing method and system that a kind of level set and deep learning combine |
| CN110189327A (en) * | 2019-04-15 | 2019-08-30 | 浙江工业大学 | Eye ground blood vessel segmentation method based on structuring random forest encoder |
Non-Patent Citations (4)
| Title |
|---|
| Hanung Adi Nugroho ; Widhia K.Z Oktoeberza ; Astrid Erasari ; Augustine Utami ; Cerwyn Cahyono."Segmentation of optic disc and optic cup in colour fundus images based on morphological reconstruction".《2017 9th International Conference on Information Technology and Electrical Engineering (ICITEE)》.2018,全文. * |
| 多相主动轮廓模型的眼底图像杯盘分割;郑姗;范慧杰;唐延东;王琰;;中国图象图形学报;20141116(第11期);第1-2节 * |
| 改进区域生长算法在视杯图像分割中的应用;刘振宇;汪淼;;辽宁大学学报(自然科学版);20170515(第02期);全文 * |
| 融合形状先验的水平集眼底图像血管分割;梁礼明;黄朝林;石霏;吴健;江弘九;陈新建;;计算机学报;20161125(第07期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110599491A (en) | 2019-12-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110599491B (en) | Priori information-based eye image segmentation method, apparatus, device and medium | |
| CN110874594B (en) | Human body appearance damage detection method and related equipment based on semantic segmentation network | |
| CN108205655B (en) | Key point prediction method and device, electronic equipment and storage medium | |
| CN108875524B (en) | Sight estimation method, device, system and storage medium | |
| JP6850399B2 (en) | Depth recovery method and equipment for monocular images, computer equipment | |
| CN110473254A (en) | A kind of position and orientation estimation method and device based on deep neural network | |
| CN110765882B (en) | Video tag determination method, device, server and storage medium | |
| CN111581414B (en) | Method, device, equipment and storage medium for identifying, classifying and searching clothes | |
| CN104781849A (en) | Fast initialization for monocular visual simultaneous localization and mapping (SLAM) | |
| CN110648289B (en) | Image noise adding processing method and device | |
| JP2020524355A5 (en) | ||
| CN109872407B (en) | Face recognition method, device and equipment, and card punching method, device and system | |
| US20220392201A1 (en) | Image feature matching method and related apparatus, device and storage medium | |
| CN110796135A (en) | Target positioning method and device, computer equipment and computer storage medium | |
| CN110866469A (en) | Human face facial features recognition method, device, equipment and medium | |
| CN111401193B (en) | Method and device for acquiring expression recognition model, and expression recognition method and device | |
| CN114219855A (en) | Point cloud normal vector estimation method and device, computer equipment and storage medium | |
| CN111860484B (en) | Region labeling method, device, equipment and storage medium | |
| CN111124902A (en) | Object operating method and device, computer-readable storage medium and electronic device | |
| CN110472537B (en) | Self-adaptive identification method, device, equipment and medium | |
| CN110427870B (en) | Eye picture recognition method, target recognition model training method and device | |
| CN112258647A (en) | Map reconstruction method and device, computer readable medium and electronic device | |
| CN109242962A (en) | The 3D display method, apparatus and electronic equipment of Lung neoplasm | |
| CN117252914A (en) | Training method and device of depth estimation network, electronic equipment and storage medium | |
| CN114373040A (en) | Three-dimensional model reconstruction method and acquisition terminal |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |